It's not that hard to project the costs of a reactor. Tokamaks for example have to be very large for net power, and the chamber has to be surrounded by neutron-shielded superconductors. That's going to be expensive.
Whether the plasma will behave the way we want is a tougher question, but roughly how much a device will cost to build is relatively easy to answer.
There's another benefit to cheaper approaches: you can afford more experiments, and you can probably do them faster. Taking a couple decades and $50 billion to build one experimental reactor is not necessarily the fastest approach to getting something to work. ITER is the ultimate waterfall project.
A lot of the cost will be ongoing operating costs, and a lot of that probably can't be easily predicted without operating experience. Of the expenses involved in running a fission plant (even in a hypothetical generous regulatory regime), how many of them could have been accurately predicted in, say, 1930?
Whether the plasma will behave the way we want is a tougher question, but roughly how much a device will cost to build is relatively easy to answer.
There's another benefit to cheaper approaches: you can afford more experiments, and you can probably do them faster. Taking a couple decades and $50 billion to build one experimental reactor is not necessarily the fastest approach to getting something to work. ITER is the ultimate waterfall project.